The tsdimaging task grids/images total power and spectral data according to a specified gridding kernel. The input data should be calibrated and bandpass corrected (where necessary), and the output is a CASA image format dataset, either 2-D, 3-D, or 4-D depending on the input parameters.

The output image contains up to four axes: two spatial axes, frequency, and polarization. By default, the spatial coordinates are determined such that the imaged area is covered with a cell size equal to 1/3 of the FWHM of the primary beam of antennas in the first MS. It is also possible to define the spatial axes of the image by specifying the image center direction (phasecenter), the number of image pixels (imsize), and the pixel size (cell).

The frequency coordinate of the image is defined by three parameters: the number of channels (nchan), the channel number/frequency/velocity of the first channel (start), and the channel width (width). The start and width parameters can be in units of 'channel' (use channel number), 'frequency' (e.g., 'GHz'), or 'velocity' (e.g., 'km/s'). By default, nchan, start, and width are set so that all selected spectral windows are covered with a channel width equal to the separation of the first two channels selected.

Finally, the polarization axis of the image is determined by the stokes parameter. For example, stokes='XXYY' produces an image cube with each plane containing the image of one of the polarizations, while stokes='I' produces a "total intensity", or Stokes I image. There is also another option for Stokes I imaging, called 'pseudoI'; the difference between 'I' and 'pseudoI' is how the task handles flag information. The stokes='I' imaging is stricter in the sense that it only takes into account visibilities for which all correlations are valid. In other words, it excludes all correlations for any data with any correlation flagged, even though the remaining correlations are valid. On the other hand, the 'pseudoI' option allows Stokes I images to include data for which either of the parallel hand data are unflagged.

NOTE: Users should set stokes='pseudoI' if you want to get the equivalent result to the one obtained by setting stokes='I' for sdimaging. Setting stokes='I' in sdimaging is implemented the same way as stokes='pseudoI' in tsdimaging.

The parameter gridfunction sets the gridding function (convolution kernel) for imaging. Currently, the task supports 'BOX' (boxcar), 'SF' (Prolate Spheroidal Wave Function), 'GAUSS' (Gaussian), 'GJINC' (Gaussian*Jinc), where Jinc(x) = $J_1(π*x/c)/(π*x/c)$ with a first order Bessel function J_1, and 'PB' (Primary Beam).

There are four subparameters for gridfunction: convsupport, truncate, gwidth, and jwidth. The convsupport parameter is an integer specifying the cutoff radius for 'SF' in units of pixels. By default (convsupport=-1), the cutoff radius is 3 pixels. The truncate parameter is a cutoff radius for 'GAUSS' or 'GJINC'. It accepts integer, float, and string values, where the string would be a number plus unit. Allowed units include 'deg', 'arcmin', 'arcsec', and 'pixel'. The default is 'pixel'. The default value for truncate, which is used when a negative radius is set, is 3*HWHM for 'GAUSS' and the radius at the first null for 'GJINC'. The gwidth is the HWHM of the Gaussian for 'GAUSS' and 'GJINC'. The default value is $sqrt(log(2))$ pixels for 'GAUSS' and $2.52*sqrt(log(2))$ pixels for 'GJINC'. The jwidth specifies the width of the jinc function (parameter 'c' in the definition above). The default is 1.55 pixels. Both gwidth and jwidth allow integer, float, or string values, where the string would be a number plus unit. The default values for gwidth and jwidth are taken from Mangum, et al. 2007 [1] . The formula for 'GAUSS' and 'GJINC' are taken from Table 1 in the paper, and are written as follows using gwidth and jwidth:

GAUSS: $\exp[-\log(2)*(|r|/gwidth)^2]$,

GJINC: $J_1(π*|r|/jwidth)/(π*|r|/jwidth)* \exp[-\log(2)*(|r|/gwidth)^2]$,

where $r$ is the radial distance from the origin. 

The imagename should be unique. tsdimaging will stop with an Exception error (e.g., Exception: Unable to open lattice) if imagename is the same as the vis name.

The ephemsrcname parameter can be set to specifiy an ephemeris for a moving source (Solar Sytem objects). If the source name in the data matches one of the Solar System objects known by CASA, the imaging realigns the data by shifting the source, so that the source appears to be fixed in the image.

The clipminmax function can clip minimum and maximum values from each pixel. This function makes the computed output slightly more robust to noise and spurious data. Note that the benefit of clipping is lost when the number of integrations contributing to each gridded pixel is small, or where the incidence of spurious data points is approximately equal to or greater than the number of beams (in area) encompassed by the expected image.

The minweight parameter defines a threshold of weight values to mask. The pixels in outfile whose weight is smaller than minweight*median (weight) are masked out. The task also creates a weight image with the name outfile.weight.

The projection parameter allows to specify what kind of map projection is applied. The default is SIN (slant orthographic) projection. The task also supports CAR (plate carrée), TAN (gnomonic), and SFL (Sanson-Flamsteed). 


Technical Note: sdimaging and tsdimaging

The tsdimaging task is supposed to replace sdimaging. The initial version of this task was intended to be fully compatible with sdimaging. Technically speaking, those tasks share underlying framework with interferometry imaging tasks: sdimaging shares with clean, while tsdimaging is based on the framework for tclean. As clean (and the underlying framework) will be deprecated and replaced with tclean in the future, sdimaging will also be made obsolete in favor of migrating to tsdimaging. This transition will have several benefits from the user's point of view in future. In terms of functionality, new features implemented in tclean will also apply to tsdimaging if the features are useful for single dish imaging. Another possible benefit is a performance. Since the framework for tclean is designed to support parallel processing, it can also be used to speed up tsdimaging. This should be effective for large datasets, but these examples represent future work. Currently, effort is underway to make tsdimaging compatible with sdimaging and convert it to a "regular" (non-experimental) task.

Citation Number 1
Citation Text Mangum, et al. 2007, A&A, 474, 679-687 (A&A)